Validation of the Louisiana Interlayer Shear Strength Test for Tack Coat (2018)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

22 The main objective of the field study was to evaluate the effects of tack coat material type, application rate, and pavement surface type, on interface bonding and pavement performance during the first 12 months of service. For this purpose, 10 field projects were selected in six states and included 33 in-service test sections. Statistical analyses were conducted on ISS test results and are reported here to ascer- tain the effects of tack coat material type, application rate, and service time for each field project. Statistical analysis software (SAS) was used. ANOVA with a least significant difference (LSD) option and a pairwise t-test were performed. A Type I error rate (α) of 0.05 was used to identify any significant differences between the tested variables. Statistical groupings obtained by each variable, such as tack coat material, appli- cation rate, and service time, were designated by letters A, B, and C. The letter A was assigned to the best performer within a group, followed by B and C. In addition, minimum and maxi- mum values of a property are represented by an error bar in figures in this chapter. 4.1 Rheological Properties of Tack Coat Materials Six types of tack coat materials—SS (SS-1H, CSS-1H and SS-1) and RS nontracking (NTSS-1HM, CBC-1H and CRS-1HBC) were evaluated on four types of pavement sur- faces; however, CRS-1HBC and SS-1 were used only on exist- ing HMA and milled HMA pavement surfaces. A suite of physical and mechanical tests was performed to measure the rheological properties of the tack coat residues. Table 4.1-1 presents the tack coat residue characterization test results. The residual asphalt content of the evaluated emulsified tack coats varied between 44% and 64%. The consistency tests performed were Saybolt Furol viscosity, penetration, softening point, and rotational viscosity. The Saybolt Furol viscosities of all tack coats ranged between 15 and 40 seconds and indi- cated that the emulsified tack coat materials were thin enough to be sprayed uniformly with the distributor. As shown in Table 4.1-1, a stiffer base asphalt binder was used for the nontracking RS NTSS-1HM tack coat; this use was evident from the rotational viscometer test results for this type of tack coat. However, nontracking RS CBC-1H and CRS-1HBC and SS SS-1H and CSS-1H tack coat resi- dues showed similar penetration values that were within the range between 40 and 71 penetration units. The SS-1 residue was softer than SS-1H according to the penetration test; see Table 4.1-1. As expected, the ranking of the softening point test results was similar to those of the penetration tests. The PG of the tack coat residues was determined according to AASHTO M 320, Standard Specification for Performance- Graded Asphalt Binder. Rheological test results of all emul- sified tack coat residues are provided in Table C-1 through Table C-8 in Appendix C. The PG for nontracking RS NTSS- 1HM tack coat residues was on the high end, because stiffer base asphalt cement was used. The CBC-1H, CRS-1HBC, SS-1H, and CSS-1H residues were graded as either PG 70-22 or PG 64-22. The SS-1 residue was graded as PG 46-28, indicating the use of relatively soft base asphalt in this type of tack coat. 4.2 Pavement Surface Texture Depths Pavement surface macrotexture depth was measured before overlay construction in all field projects according to ASTM E965, Standard Test Method for Measuring Pavement Macro- texture Depth Using a Volumetric Technique, a test commonly referred to as the sand patch test. Four types of pavement surfaces—new HMA, existing HMA, milled HMA, and PCC— were evaluated. Three measurements were taken within each test section. Table 4.2-1 presents the mean surface texture depths (MTD) for each project including the averaged MTD for different pavement surface types. As shown in the table, the milled HMA surface exhibited the highest average MTD, C H A P T E R 4 Results and Analysis

23 Project Tack Coat Material Type AASHTO Test Method T 59 T 59 T 49 T 53 M 320 Residue by Evaporation, % Viscosity, Saybolt Furol at 25°C, s Penetration at 25°C, 100g, 5s, penetration units Softening Point, Ring and Ball, °C Superpave PG Missouri SS-1H1 61.0 29.2 71.0 51.4 PG 64-22 NTSS-1HM2 63.0 41.5 9.0 82.0 PG 94-10 Louisiana (LA30) SS-11 64.1 32.7 102.0 43.5 PG 46-28 NTSS-1HM2 54.3 34.2 9.0 78.1 PG 82-10 Louisiana (LA1053) NTSS-1HM2 43.6 16.0 8.0 78.3 PG 94-4 CBC-1H2 51.7 15.2 40.3 56.4 PG 70-16 NTSS-1HM2 57.1 16.2 8.7 72.5 PG 88-10 SS-1H1 57.8 25.1 45.3 55.8 PG 70-22 Florida SS-1H1 60.0 23.5 50.3 52.5 PG 64-22 CRS-1HBC2 59.3 19.5 68.4 50.2 PG 64-22 Tennessee NTSS-1HM2 52.3 16.7 8.0 79.2 PG 100-10 CBC-1H2 52.5 15.2 48.3 55.1 PG 70-22 CSS-1H1 61.5 23.3 66.3 52.6 PG 64-22 Nevada CBC-1H2 59.1 18.0 58.3 52.0 PG 70-28 CSS-1H1 48.1 16.4 53.0 52.2 PG 70-22 Oklahoma CBC-1H2 51.2 17.8 52.7 55.0 PG 64-22 CSS-1H1 61.7 38.2 53.0 51.0 PG 64-22 Note: Penetration units = 0.1 mm. 1SS. 2Nontracking RS tack coat material. ® Table 4.1-1. Characterization of emulsified tack coat residues: all projects. Pavement Surface Type Project Surface MTD, mm Mean Surface MTD, mm Range, mm Milled HMA Missouri 1.62 1.77 1.38–2.14 Louisiana 1.56 Tennessee 1.92 Nevada 1.83 New HMA Missouri 0.87 0.91 0.84–0.95 Louisiana 0.93 Existing HMA Missouri 0.99 0.97 0.95–0.99 Florida 0.96 PCC Missouri 1.26 1.49 1.25–1.67 Oklahoma 1.61 Note: 1 mm = 0.04 in. Table 4.2-1. Surface MTDs: all projects.

24 followed by PCC, existing HMA, and new HMA. The range of measured MTD values for each surface type is presented in Table 4.2-1. 4.3 ISS Test Results The ISS test was conducted at a temperature of 25°C, in accordance with AASHTO TP 114, Standard Method of Test for Determining the Interlayer Shear Strength (ISS) of Asphalt Pavement Layers. Triplicate samples were tested for each con- dition. A summary of ISS test results for all projects, includ- ing the average, standard deviation, and COV, is presented in Table 4.3-1 through Table 4.3-8. The COV for ISS was less than 15% on average for samples tested. Observed trends and effects of the design variables are discussed in the follow- ing sections. Pavement Surface Type ISS, psi NTSS-1HM SS-1H 0.05 gsy1 0.05 gsy 0M 7M 12M 0M 7M 12M Milled HMA AVG 76 80 85 75 77 88 STD 12 10 12 13 3 15 COV, % 16 13 14 17 4 17 New HMA AVG 59 94 122 40 46 56 STD 2 3 5 5 3 1 COV, % 3 3 4 13 7 2 Existing HMA AVG 40 48 52 28 33 37 STD 8 4 5 2 2 3 COV, % 19 7 9 8 8 8 PCC AVG 23 28 40 19 22 29 STD 3 3 8 3 3 3 COV, % 13 11 20 16 14 10 Note: 0M, 7M, and 12M indicate ISS at 0, 7, and 12 months, respectively; 1 psi = 6.90 kpa. 1 Residual application rate. Table 4.3-1. ISS test results: Missouri project. Pavement Surface Type ISS, psi NTSS-1HM SS-1 0.06 gsy1 0.06 gsy 0M 4M 12M 0M 4M 12M Milled HMA AVG 80 82 118 38 37 71 STD 10 8 9 5 6 2 COV, % 13 10 8 13 16 3 Note: 0M, 4M, and 12M indicate ISS at 0, 4, and 12 months, respectively. 1 Residual application rate. Table 4.3-2. ISS test results: Louisiana project (LA 30). Pavement Surface Type ISS, psi NTSS-1HM1 CBC-1H 0.01 gsy2 0.02 gsy 0.02 gsy 0.04 gsy 0M 4M 12M 0M 4M 12M 0M 4M 12M 0M 12M New HMA AVG 55 56 55 80 82 85 41 71 97 66 93 115 STD 6 8 6 6 7 11 5 10 13 5 12 10 COV, % 11 14 11 8 9 13 12 14 13 8 13 9 Note: 0M, 4M, and 12M indicate ISS at 0, 4, and 12 months, respectively. 1Blacklidge nontracking RS emulsion. 2Residual application rate. 4M Table 4.3-3. ISS test results: Louisiana project (LA 1053).

25 Pavement Surface Type ISS, psi NTSS-1HM1 SS-1H 0.02 gsy2 0.03 gsy 0.02 gsy 0.03 gsy 0M 4M 12M 0M 4M 12M 0M2 4M 12M 0M 4M 12M New HMA AVG 68 93 120 76 105 110 52 80 102 58 90 96 STD 8 12 9 4 7 13 3 6 10 6 4 4 COV, % 12 13 8 5 7 12 6 8 10 10 4 4 Note: 0M, 4M, and 12M indicate ISS at 0, 4, and 12 months, respectively. 1Asphalt Products Unlimited nontracking RS emulsion. 2Residual application rate. Table 4.3-4. ISS test results: Louisiana project (LA 1053). Pavement Surface Type ISS, psi CRS-1HBC SS-1H 0.02 gsy1 0.04 gsy 0.04 gsy 0.04 gsy 0M 4M 12M 0M 4M 12M 0M 4M 12M 0M 4M 12M Existing HMA AVG 26 56 56 46 69 74 44 62 86 43 58 68 STD 2 3 5 2 7 1 2 7 11 1 2 8 COV, % 8 5 9 4 10 1 5 11 13 2 3 11 Note: 0M, 4M, and 12M indicate ISS at 0, 4, and 12 months, respectively. 1Residual application rate. Table 4.3-5. ISS test results: Florida project. Pavement Surface Type ISS, psi NTSS-1HM CBC-1H CSS-1H 0.05 gsy1 0.06 gsy 0.05 gsy 3M 12M 3M 12M 3M 12M Milled HMA AVG 126 133 106 127 96 107 STD 17 11 14 11 9 12 COV, % 13 8 13 9 9 11 Note: 3M and 12M indicate ISS at 3, and 12 months, respectively. 1Residual application rate. Table 4.3-6. ISS test results: Tennessee project. Pavement Surface Type ISS, psi CBC-1H CSS-1H 0.03 gsy1 0.04 gsy 0.05 gsy 0.07 gsy 0M 10M 12M 0M 10M 12M 0M 10M 12M 0M 10M 12M Milled HMA AVG 84 133 125 88 147 143 94 108 110 99 139 143 STD 10 13 5 11 15 20 10 14 9 7 9 13 COV, % 12 10 4 13 10 14 11 13 8 7 6 9 Note: 0M, 10M, and 12M indicate ISS at 0, 10, and 12 months, respectively. 1Residual application rate. Table 4.3-7. ISS test results: Nevada project.

26 4.4 Effect of Tack Coat Material Type on ISS 4.4.1 Missouri Project Figure 4.4-1 presents the effect of two types of tack coat materials: nontracking RS NTSS-1HM and SS SS-1H. Both tack coat materials were sprayed at a 0.05 gsy residual appli- cation rate on four types of pavement surfaces: milled HMA, new HMA, existing HMA and PCC. The horizontal dashed line illustrates the minimum recommended ISS threshold value (1). As shown in Table 4.3-1, the nontracking NTSS- 1HM tack coat provided higher ISS values than the SS-1H tack coat at 0-month service (i.e., immediately after overlay construction) for all types of pavement surfaces when other factors were held constant; the result may be attributed to the high stiffness or viscosity of the nontracking NTSS-1HM tack coat residue, Table 4.1-1. However, the differences between ISS values were not statistically significant for all types of pave- ment surfaces except for new HMA when NTSS-1HM was compared with SS-1H (Figure 4.4-1). 4.4.2 Louisiana Project Figure 4.4-2 presents the effect of two types of emulsi- fied tack coat materials—nontracking SS NTSS-1HM and SS SS-1—on ISS at 0-month service. Both tack coat materials were sprayed at 0.06 gsy residual application rate on a milled HMA pavement surface. The SS-1 tack coat did not meet the minimum recommended ISS threshold value of 40 psi at 0-month service (Table 4.3-2). The nontracking NTSS-1HM provided significantly greater bonding strength than SS-1; this result can be attributed to the fact that the base asphalt of the NTSS-1HM (PG 82-10) was much stiffer than that for SS-1 (PG 46-28) (Table 4.1-1). Pavement Surface Type ISS, psi CBC-1H CSS-1H 0.03 gsy1 0.04 gsy 0.07 gsy 0.08 gsy 0M 4M 12M 0M 4M 12M 0M 4M 12M 0M 4M 12M PCC AVG 44 52 61 55 84 96 50 66 88 58 83 90 STD 6 9 6 9 10 17 4 9 14 8 12 9 COV, % 14 17 10 16 12 18 8 14 16 14 14 10 Note: 0M, 4M, and 12M indicate ISS at 0, 4, and 12 months, respectively. 1Residual application rate. Table 4.3-8. ISS test results: Oklahoma project. 0 40 80 120 160 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy NTSS-1HM SS-1H NTSS-1HM SS-1H NTSS-1HM SS-1H NTSS-1HM SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A A A A A A A B Milled HMA New HMA Existing HMA PCC Figure 4.4-1. Effect of tack coat material type on ISS: Missouri project. (A and B denote whether the difference between the values represented by the two columns is statistically significant. When both columns are labeled A, the difference is not statistically significant.)

27 Figure 4.4-3 shows the effect of four types of tack coat materials—three nontracking RS (two NTSS-1HM and one CBC-1H) and one SS SS-1H—on ISS on a new HMA pavement surface. Both NTSS-1HM tack coats provided the highest ISS values, followed by SS-1H and CBC-1H, when other factors were kept constant, as shown in Table 4.3-3. The mean ISS values of both NTSS-1HM materials were statistically greater than CBC-1H and SS-1H when compared at a 0.02 gsy residual application rate. As noted earlier, this result may be due to the high stiffness, viscosity, and softening point of the NTSS-1HM residues, Table 4.1-1. Further, the statistically similar perfor- mance of CBC-1H (PG 70-16) and SS-1H (PG 70-22) tack coat materials may be attributed to the fact that both tack coat resi- dues had comparable rheological properties; see Table 4.1-1. 4.4.3 Florida Project This project evaluated two types of tack coat materials— nontracking RS CRS-1 HBC and SS SS-1H—on an existing HMA pavement surface. Figure 4.4-4 presents the effect of tack coat materials on ISS at 0-month service. Both tack coat materials provided statistically equivalent ISS when com- pared at 0.04 gsy residual application rate (Table 4.3-5). This result may be attributed to the similar rheological properties of both CRS-1 HBC (PG 64-22) and SS-1H (PG 64-22) residue (Table 4.1-1). 4.4.4 Tennessee Project This project evaluated three types of tack coat materials: two nontracking RS (CBC-1H and NTSS-1HM) and an SS CSS-1H on a milled HMA pavement surface. Figure 4.4-5 presents the effect of tack coat materials on ISS at 3 months in service. All three materials exhibited significantly higher ISS values than the minimum recommended ISS threshold value (1); see Table 4.3-6. This performance may be attrib- uted to the combined effect of high residual application rate and high surface MTD; see Table 4.2-1. The nontracking NTSS-1HM provided the highest ISS of the three tack coats when other factors were kept constant; however, the effect of tack coat type on ISS was not statistically different; see Figure 4.4-5. This outcome may be due to the high stiffness, viscosity, and softening point of the NTSS-1HM (PG 100-10) 0 40 80 120 160 0.06 gsy 0.06 gsy NTSS-1HM SS-1 In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A B Milled HMA Figure 4.4-2. Effect of tack coat material type on ISS: Louisiana project (LA 30). 0 40 80 120 160 0.02 gsy NTSS-1HM 0.02 gsy CBC-1H 0.02 gsy NTSS-1HM 0.02 gsy SS-1H In te rf ac e S he ar S tr en gt h, p si Tack Coat Residual Application Rate/ Material Type New HMA A B A B Figure 4.4-3. Effect of tack coat material type on ISS: Louisiana project (LA 1053) (NTSS-1HM: PG 94-4 residual binder; NTSS-1HM: PT 88-10 residual binder).

28 residue compared with CBC-1H (PG 70-22) and CSS-1H (PG 64-22) residues (Table 4.1-1). 4.4.5 Nevada Project Two types of tack coat materials—nontracking RS CBC-1H and SS CSS-1H—were evaluated on a milled HMA pave- ment surface. Figure 4.4-6 presents the effect of tack coat materials on ISS at 0-month service. Both tack coat materials exhibited significantly higher ISS values than the minimum recommended ISS threshold value (1), see Table 4.3-7. Both CSS-1H and CBC-1H provided similar (i.e., not statisti- cally different) mean ISS values when comparing CSS-1H at 0.05 gsy with CBC-1H at 0.04 gsy residual application rates; see Figure 4.4-6. The similar performance of CBC-1H (PG 70-28) and CSS-1H (PG 70-22) may be attributed to the fact that both tack coat residues had comparable rheological properties; see Table 4.1-1. 4.4.6 Oklahoma Project This project evaluated the effect of two types of tack coat materials (nontracking RS CBC-1H and SS CSS-1H) on a PCC pavement surface. Figure 4.4-7 presents the effect of tack coat materials on ISS at 0-month service. Both tack coat materials met the minimum recommended ISS threshold value (1); see Table 4.3-8. Both tack coats provided similar (i.e., not statistically different) mean ISS values when com- 0 40 80 120 160 0.04 gsy CRS-1 HBC 0.04 gsy SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A A Existing HMA Figure 4.4-4. Effect of tack coat material type on ISS: Florida project. 0 40 80 120 160 0.05 gsy 0.06 gsy 0.05 gsy NTSS-1HM CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type Milled HMA A A A Figure 4.4-5. Effect of tack coat material type on ISS: Tennessee project. 0 40 80 120 160 0.04 gsy 0.05 gsy CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A A Milled HMA Figure 4.4-6. Effect of tack coat material type on ISS: Nevada project.

29 paring CSS-1H at 0.07 gsy with CBC-1H at 0.04 gsy residual application rates. Similar performance of both materials may be attributed to the fact that both CBC-1H (PG 64-22) and CSS-1H (PG 64-22) tack coat residues had comparable rheo- logical properties; see Table 4.1-1. 4.4.7 Summary of the Effect of Tack Coat Material Type on ISS This study evaluated the effect of six types of emulsified tack coat materials—SS (SS-1H, CSS-1H, and SS-1) and nontracking RS (CBC-1H, CRS-1HBC, and NTSS-1HM)— on four pavement surface types; however, CRS-1HBC and SS-1 tack coat materials were evaluated only on existing HMA and milled HMA surfaces. With respect to the effect of tack coat material type on ISS, nontracking RS NTSS-1HM tack coats exhibited the highest ISS on all surface types. This per- formance may be attributed to the stiffer base asphalt cement used in nontracking RS NTSS-1HM tack coat; see Table 4.1-1. However, nontracking RS CBC-1H and CRS-1HBC tack coats showed similar interface bonding in all projects when compared with SS SS-1H and CSS-1H tack coats because of similar rheological properties of the residual asphalt binders. Furthermore, SS SS-1 tack coat material resulted in the lowest ISS on a milled HMA pavement surface. The soft base asphalt cement used in SS-1 (PG 46-28) tack coat material may be the reason for the lower interface bonding. 4.5 Effect of Pavement Surface Type on ISS This study evaluated the effect of four types of pavement surfaces (milled HMA, new HMA, existing HMA, and PCC) on interface bonding between pavement layers. Figure 4.5-1 illustrates the change in ISS immediately after overlay con- struction with pavement surface types in all projects. On average, the milled HMA surface provided the highest ISS followed by new HMA, existing HMA, and PCC surface types. As shown in Table 4.2-1, the milled HMA surface had the highest MTD, which appears to be a leading factor contributing to the higher mean ISS value. However, the SS-1 tack coat material in the milled HMA exhibited an ISS value slightly below the minimum recommended ISS thresh- old value of 40 psi; see Figure 4.5-1. This observation may be attributed to the use of soft base asphalt cement in the SS-1 (PG 46-28) tack coat material (Table 4.1-1). The existing HMA and PCC surface types showed similar performance in terms 0 40 80 120 160 0.04 gsy 0.07 gsy CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A A PCC Figure 4.4-7. Effect of tack coat material type on ISS: Oklahoma project. 0 40 80 120 160 In te rf ac e Sh ea r St re ng th , p si MO FL NV LA 30 TN LA 1053 OK AVG Milled HMA New HMA PCCExisting HMA SS-1 Figure 4.5-1. Effect of pavement surface type on ISS: all projects.

30 of ISS. It was observed that the PCC surface in Missouri project exhibited the weakest interface bonding, yielding ISS values below the minimum recommended ISS threshold of 40 psi (1). Inadequate pavement surface conditions before the overlay and variability during construction could be a leading reason for such performance. Moreover, the leveling layer in Louisiana (LA 30) and Tennessee projects had minimal influ- ence on tack coat performance. During bond strength testing, it was observed that the leveling course layer became part of the overlay; thus, failure occurred at the milled surface inter- face for those two projects. 4.6 Effect of Residual Application Rate on ISS 4.6.1 Louisiana Project The Louisiana LA1053 project evaluated the effect of application rates for four types of tack coat materials—three nontracking RS (two NTSS-1HM and one CBC-1H) and one SS SS-1H—on a new HMA pavement surface type. Each type of tack coat material was evaluated at two application rates. Figure 4.6-1 presents the effect of residual application rates on the mean ISS values at 0-month service. All residual application rates in this project yielded ISS values greater than the minimum recommended ISS threshold of 40 psi (1); see Table 4.3-4. In general, mean ISS values increased with the increase in tack coat residual application rate for all types of tack coat materials. Furthermore, the increase of mean ISS values with the increase in the residual application rate was statistically significant for NTSS-1HM (Blacklidge) and CBC-1H tack coat materials; see Figure 4.6-1. 4.6.2 Florida Project Figure 4.6-2 presents the effect of residual application rate on the mean ISS values for an existing HMA pavement sur- face type at 0-month service. Both SS SS-1H and nontrack- ing RS CRS-1HBC tack coat sections exceeded the minimum ISS threshold value of 40 psi at a residual application rate of 0.04 gsy (Figure 4.6-2). As expected, both SS-1H tack coat sections at 0.04 gsy residual application rates showed similar (i.e., not statistically different) ISS values. In addition, the nontracking CRS-1HBC tack coat at 0.02 gsy did not meet the minimum ISS threshold (1), but the mean ISS value increased significantly at a residual application rate of 0.04 gsy (Figure 4.6-2). 4.6.3 Nevada Project Figure 4.6-3 presents the effect of application rate on ISS for two types of tack coat materials—nontracking RS CBC-1H and SS CSS-1H—on a milled HMA pavement surface at 0-month service. All ISS values were significantly higher than the mini- mum recommended ISS threshold of 40 psi (1); see Table 4.3-7. Further, the mean ISS values increased with an increase in the residual application rate; however, this increase was not statis- tically significant for both materials; see Figure 4.6-3. 4.6.4 Oklahoma Project Figure 4.6-4 presents the effect of residual application rates on ISS on a PCC pavement surface at 0-month service. The mean ISS values for the materials evaluated were greater than the minimum recommended ISS threshold value of 0 40 80 120 160 0.01 gsy 0.02 gsy 0.02 gsy 0.04 gsy 0.02 gsy 0.03 gsy 0.02 gsy 0.03 gsy NTSS-1HM CBC-1H NTSS-1HM SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A B A A New HMA A B A A Figure 4.6-1. Effect of residual application rate on ISS: Louisiana project (LA 1053) (NTSS-1HM, left, PG 94-4 residual binder, and NTSS-1HM, center right, PG 88-10 residual binder).

0 40 80 120 160 0.02 gsy 0.04 gsy 0.04 gsy 0.04 gsy CRS-1HBC SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A B A A Existing HMA Figure 4.6-2. Effect of residual application rate on ISS: Florida project. 0 40 80 120 160 0.03 gsy 0.04 gsy 0.05 gsy 0.07 gsy CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type A A A A Milled HMA Figure 4.6-3. Effect of residual application rate on ISS: Nevada project. 0 40 80 120 160 0.03 gsy 0.04 gsy 0.07 gsy 0.08 gsy CBC-1H CSS-1H In te rf ac e S he ar S tr en gt h, p si Tack Coat Residual Application Rate/ Material Type A A A A PCC Figure 4.6-4. Effect of residual application rate on ISS: Oklahoma project.

32 40 psi (Table 4.3-8). In general, the mean ISS values increased slightly with an increase in residual application rate; however, this increase was not statistically significant for both CBC-1H and CSS-1H tack coat materials; see Figure 4.6-4. 4.6.5 Summary of the Effect of Residual Application Rate on ISS Within the evaluated residual application rate range, the mean ISS increased with an increase in the residual application rate for all tack coat types and for all pavement surface types. Furthermore, the effect of residual application rate on ISS was found to be statistically significant for new HMA and existing HMA surface types. 4.7 Effect of Service Time on ISS 4.7.1 Missouri Project To determine the effect of service time on ISS, specimens were cored from the right wheel path of each test section at 0, 7, and 12 months of service. Although the original sched- ule was to obtain the cores at an intermediate service time of 4 months, coring was delayed to 7 months because of winter weather limitations. Figure 4.7-1 shows the variation of the mean ISS values at different service times. In general, the mean ISS increased with service time for both tack coat materials and for all surface types as a result of tack coat curing; see Table 4.3-1. Similar observations were reported by Sholar et al. (26) and Hachiya and Sato (24). Furthermore, LSD tests on the ISS test results at the three service times revealed that interface bonding was not statistically different for all surface types except for the nontracking RS NTSS-1HM tack coat section on new HMA surface. Hence, service time had a sig- nificant effect in increasing the interface bonding strength for the combination of new HMA surface and the nontracking RS NTSS-1HM tack coat. 4.7.2 Louisiana Project For the Louisiana LA 30 project, roadway cores were obtained at service times of 0, 4, and 12 months. Figure 4.7-2 shows the variation of the mean ISS values at different service times on a milled HMA pavement surface. SS-1 did not meet the minimum recommended ISS threshold value (1) at 0 and 4 months of service; see Table 4.3-2. Statistical analysis on the ISS results revealed that interface bonding at 12 months of service was significantly higher than that at 0 and 4 months for both NTSS-1HM and SS-1 tack coats. Hence, service time had a pronounced effect for both tack coat types. For the LA 1053 project, specimens were obtained at 0, 4, and 12 months of service to determine the effect of service time on ISS on a new HMA pavement surface. Figure 4.7-3 shows the variation of the mean ISS values at different service times. All mean ISS values for the tack coat materials evalu- ated were greater than the recommended ISS threshold value of 40 psi (1); see Tables 4.3-3 and 4.3-4. The mean ISS values increased with service time because of the tack coat curing effect (24, 26). Statistical analyses on the ISS results at the three service times showed statistical difference for all cases except for the NTSS-1HM (Blacklidge) sections; this differ- ence suggests that the curing effect is less pronounced for nontracking RS NTSS-1HM (Blacklidge) tack coat material; see Figure 4.7-3. 0 40 80 120 160 200 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy NTSS-1HM SS-1H NTSS-1HM SS-1H NTSS-1HM 0.05 gsy SS-1H NTSS-1HM SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 0M 7M 12M A AA A A A A B C A AA AAA AA A A AA AAA Milled HMA New HMA Existing HMA PCC Figure 4.7-1. Effect of service time on ISS: Missouri project. (If three columns are labeled A, B, and C, differences in their mean values are all statistically significant.)

33 4.7.3 Florida Project This project evaluated the effect of two types of tack coat materials at two residual application rates on ISS at service times of 0, 4, and 12 months on an existing HMA pavement surface. Figure 4.7-4 shows the variation of the mean ISS val- ues at different service times. In general, the mean ISS values increased with service time for both tack coat materials at both residual application rates; see Table 4.3-5. Statistical analyses indicated significantly different mean ISS values at different service times for both tack coat materials. The SS-1H sections exhibited a considerable improvement in interface bonding strength as compared with the nontracking CRS- 1HBC test sections; the difference suggests that the long-term curing effect may be more pronounced for SS SS-1H than for nontracking RS CRS-1HBC; see Figure 4.7-4. 4.7.4 Tennessee Project Specimens were obtained from the right wheel path of each test section at 3 and 12 months of service. Although the original testing plan was to collect the first set of cores at 0-month service, problems associated with coring rigs necessitated collection of the first set of cores at 3 months of service. Hence, only two service times were considered. Figure 4.7-5 presents the effect of service time on ISS on a 0 40 80 120 160 200 0.06 gsy 0.06 gsy NTSS-1HM SS-1 In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 0M 4M 12M B B A A BB Milled HMA Figure 4.7-2. Effect of service time on ISS: Louisiana project (LA 30). 0 40 80 120 160 200 0.01 gsy 0.02 gsy 0.02 gsy 0.04 gsy 0.02 gsy 0.03 gsy 0.02 gsy 0.03 gsy NTSS-1HM¹ CBC-1H NTSS-1HM² SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 0M 4M 12MNew HMA A A A A A A A B C A A B A B C A A B A B C A B A Figure 4.7-3. Effect of service time on ISS: Louisiana project (LA 1053) (NTSS-1HM, left, PG 94-4, residual binder, and NTSS-1HM, center right, PG 88-10 residual binder).

34 milled HMA pavement surface. All mean ISS values were significantly higher than the minimum recommended ISS threshold value of 40 psi; see Table 4.3-6. The mean ISS values increased with service time because of tack coat curing (24, 26). However, LSD tests indicated that this increase was not statistically significant; see Figure 4.7-5. 4.7.5 Nevada Project This project evaluated the effect of two types of tack coat materials at two residual application rates on ISS at service times of 0, 10, and 12 months on a milled HMA pavement surface. Although the original testing plan was to collect the second set of cores at 4 months of service, logistic issues related to the coring equipment required collection of the second set of cores at 10 months of service. Figure 4.7-6 presents the effect of service time on ISS on a milled HMA pavement surface. All mean ISS values were higher than the minimum recommended ISS threshold value of 40 psi; see Table 4.3-7. As stated earlier, the high surface roughness of the milled HMA pavement was a major contributing factor to the high ISS values observed; see Table 4.2-1. Regarding the effect of service time, all mean ISS values increased with an increase in service time because of the curing effect (24, 26). Results of LSD tests on the ISS test results at the three service times showed statistical significance for all cases except for the CSS-1H 0 40 80 120 160 200 0.02 gsy 0.04 gsy 0.04 gsy 0.04 gsy CRS-1HBC SS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 0M 4M 12M A A B AA B A B C A A B Existing HMA Figure 4.7-4. Effect of service time on ISS: Florida project. 0 40 80 120 160 200 0.05 gsy 0.06 gsy 0.05 gsy NTSS-1HM CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 3M 12M AA A A A A Milled HMA Figure 4.7-5. Effect of service time on ISS: Tennessee project.

35 at 0.05 gsy; see Figure 4.7-6. Furthermore, the curing effect seemed more pronounced for the RS nontracking CBC-1H than the SS CSS-1H. 4.7.6 Oklahoma Project This project evaluated the effect of two types of tack coat materials at two residual application rates on ISS at service times of 0, 4, and 12 months. Figure 4.7-7 presents the effect of service time on ISS on a grooved PCC pavement surface. All mean ISS values were higher than the minimum recom- mended ISS threshold value of 40 psi; see Table 4.3-8. In gen- eral, all mean ISS values increased with service time because of curing (24, 26). Statistical analyses on the ISS values at the three service times showed statistical difference for all cases except for the CBC-1H at 0.03 gsy; see Figure 4.7-7. Hence, the curing effect seemed to be more pronounced with the use of CSS-1H at a relatively high residual application rate. 4.7.7 Summary of the Effect of Service Time on ISS Regarding the effect of service time on ISS, it was observed that interface bonding strength increased with service time in all field projects and for all pavement surface types. This phenomenon was primarily attributed to tack coat curing and seemed more pronounced for the combination of SS tack coat materials and new HMA pavement surface type. Further, 0 40 80 120 160 200 0.03 gsy 0.04 gsy 0.05 gsy 0.07 gsy CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 0M 10M 12MMilled HMA A A B AA B AA A AA B Figure 4.7-6. Effect of service time on ISS: Nevada project. 0 40 80 120 160 200 0.03 gsy 0.04 gsy 0.07 gsy 0.08 gsy CBC-1H CSS-1H In te rf ac e Sh ea r St re ng th , p si Tack Coat Residual Application Rate/ Material Type 0M 4M 12MPCC A A A A A B A A B A A B Figure 4.7-7. Effect of service time on ISS: Oklahoma project.

36 the curing effect of tack coat materials seemed to increase with the increase in the residual application rate. 4.8 Analysis of FWD Test Results 4.8.1 Missouri Project A series of FWD tests was performed along the pavement sections before overlay construction and after construction at 7 months and 12 months of service. Figure 4.8-1 presents the change in average center deflections with service time for the different pavement surface types considered. In general, the average center deflections for all surfaces decreased after overlay construction because of the added structural layer. Furthermore, a slight decrease in deflections was noted with service time and may be attributed to densification of the HMA overlays due to in-service trafficking and improved interface bonding with service time (Figure 4.7-1). The aver- age center deflections for the existing HMA surface were sig- nificantly higher compared with the other surface types; this difference may indicate poor structural conditions in these test sections. However, the decreases in center deflections with service time on the existing HMA surface were notice- ably higher compared with other surfaces. 4.8.2 Louisiana Project In the Louisiana LA 30 project, a series of FWD tests was performed at 4 months and 12 months of service time. Although the original experimental plan was to perform FWD testing before and after construction of overlays, prob- lems associated with construction necessitated FWD testing at 4 months of service. Figure 4.8-2 presents the change in average center deflections with service time. The average center deflections decreased slightly with service time. Densification of HMA overlays due to in-service trafficking and improved interface bonding due to tack coat curing (Figure 4.7-2) could be a reason for the decrease in FWD center deflection. Densi- fication was investigated through air voids measurements in the Oklahoma project. Figure 4.8-3 presents the change in FWD center deflections with service time for the Louisiana LA 1053 project. A series of FWD tests was performed along the pavement sections before HMA overlay construction and after construction at 4 months and 12 months of service. In general, a decrease in average center deflections occurred in all test sections after placement of HMA overlays because of added structural layer. 0 5 10 15 20 25 30 35 40 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy 0.05 gsy NTSS-1HM SS-1H NTSS-1HM SS-1H NTSS-1HM SS-1H NTSS-1HM 0.05 gsy SS-1H FW D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Material Type Pre-overlay 7M 12M Existing HMA Milled HMA New HMA PCC Figure 4.8-1. Comparison of FWD center deflections: Missouri project. 0 5 10 15 20 25 30 35 40 0.06 gsy 0.06 gsy NTSS-1HM SS-1 F W D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Material Type 4M 12MMilled HMA Figure 4.8-2. Comparison of FWD center deflections: Louisiana project (LA 30).

37 Further, a slight decrease in FWD deflections occurred in all but two test sections [i.e., NTSS-1HM (Blacklidge) at 0.01 gsy, and SS-1H test sections at 0.02 gsy]. This decrease in deflec- tion could be because of the densification of overlays and improved interface bonding; see Figure 4.7-3. 4.8.3 Florida Project Figure 4.8-4 presents a comparison of the average FWD center deflections with service time for Florida project. A series of FWD tests was performed along the test sections before overlay construction, and after construction at 4 months and 12 months of service. As shown in the figure, average center deflections decreased significantly after overlay con- struction because of the added structural layer. Moreover, a slight decrease in deflection occurring with service time can be attributed to the densification of HMA overlays due to in-service trafficking and an increase in interface bonding with time; see Figure 4.7-4. 4.8.4 Tennessee Project Figure 4.8-5 presents the change in average FWD center deflections with time for the Tennessee project. A series of FWD tests was performed at 3 months and 12 months of ser- vice. Although the original experimental plan was to perform 0 5 10 15 20 25 30 35 40 0.01 gsy 0.02 gsy 0.02 gsy 0.04 gsy 0.02 gsy 0.03 gsy 0.02 gsy 0.03 gsy NTSS-1HM CBC-1H NTSS-1HM SS-1H F W D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Material Type Pre-overlay 4M 12MNew HMA Figure 4.8-3. Comparison of FWD center deflections: Louisiana project (LA 1053) (NTSS-1HM, left, PG 94-4 residual binder, and NTSS-1HM, center right, PG 88-10 residual binder). 0 5 10 15 20 25 30 35 40 0.02 gsy CRS-1HBC 0.04 gsy CRS-1HBC 0.04 gsy SS-1H 0.04 gsy SS-1H F W D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Material Type Pre-overlay 4M 12MExisting HMA Figure 4.8-4. Comparison of FWD center deflections: Florida project.

38 FWD testing before and after construction of overlays, prob- lems associated with construction difficulties allowed FWD testing at 3 months of service. The average center deflec- tions for all test sections decreased with time because of densification of HMA overlays and improved interface bond- ing; see Figure 4.7-5. 4.8.5 Nevada Project Figure 4.8-6 presents the change in average FWD deflec- tions with time for the Nevada project. As shown in the fig- ure, the average center deflections at 0 month and 4 months of service were the same; however, a slight decrease in deflec- tion was observed at 12 months of service. Densification of overlays due to trafficking and improved interface bonding (Figure 4.7-6) could be a reason for this decrease. 4.8.6 Oklahoma Project Figure 4.8-7 presents the effect of service time on air voids in the Oklahoma project. Because of in-service trafficking, air voids decreased with service time for all test sections eval- uated. Further, a series of FWD tests was performed before overlay construction, and after construction at 4 months of service on two test sections (i.e., CBC-1H at 0.03 gsy and CSS-1H at 0.08 gsy). As shown in Figure 4.8-8, a slight 0 5 10 15 20 25 30 35 40 0.05 gsy NTSS-1HM 0.06 gsy CBC-1H 0.05 gsy CSS-1H F W D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Material Type 3M 12MMilled HMA Figure 4.8-5. Comparison of FWD center deflections: Tennessee project. 0 5 10 15 20 25 30 35 40 0.03 gsy 0.04 gsy 0.05 gsy 0.07 gsy CBC-1H CSS-1H F W D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Material Type 0M 4M 12M Figure 4.8-6. Comparison of FWD center deflections: Nevada project.

39 decrease in the average center deflections occurred after over- lay construction because of the combined effect of added structural layer, densification due to trafficking, and improved interface bonding due to tack coat curing. 4.8.7 Summary of FWD Test Results Mean FWD center deflections decreased with service time in all field projects and for all pavement surface types. This phenomenon was primarily attributed to the densification effect of HMA overlays due to in-service trafficking and improved interface bonding with service time. Air void mea- surements in the Oklahoma project validated the densification of the overlay with time. 4.9 Relationship Between ISS and FWD Center Deflection Figure 4.9-1 presents a relationship between ISS test results and average FWD center deflection measured in all field projects. As shown in the figure, a decreasing trend of ISS was observed with increasing average FWD center deflection. This trend suggests that the structural capacity of pavements depends on the interface bonding between 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 0.03 gsy 0.04 gsy 0.07 gsy 0.08 gsy CBC-1H CSS-1H A ir V oi d, % 0M 4M 12M Figure 4.8-7. Density test results: Oklahoma project. 0 5 10 15 20 25 30 35 40 0.03 gsy CBC-1H 0.08 gsy CSS-1H F W D C en te r D ef le ct io n, m ill s Tack Coat Residual Application Rate/ Materials Type Pre-overlay 4MPCC Figure 4.8-8. Comparison of FWD center deflections: Oklahoma project.

40 pavement layers. Further, poor interface bonding (<40 psi) can be noted even though the average FWD center deflections were on the lower end (<10 mils). This result suggests that factors affecting interface bonding between pavement layers (i.e., tack coat material type, residual application rate, pave- ment surface condition, construction practice, and pavement temperature) may also influence such performance. 4.10 Short-Term Performance of Test Sections Figure 4.10-1 presents a histogram showing all individual ISS measurements from all projects including different ser- vice times. The ISS results ranged between 0 and 180 psi. An ISS value of zero indicates an already fractured interface during coring, while on the upper end, the ISS value of 180 psi was obtained from milled HMA test sections after 12 months of trafficking. In total, 320 samples were tested, and the aver- age ISS value was 76 psi with a standard deviation of 32 psi. As shown in the figure, approximately 72 percent of ISS mea- surements were between 40 and 120 psi, which is within one standard deviation of the mean ISS. Approximately 12 per- cent of ISS values did not meet the minimum recommended ISS threshold (1). As expected, most test sections evaluated, except for a few sections in the Missouri project, performed satisfactorily in the field; see Figure 4.10-2. A pavement distress survey was performed over the entire length of each pavement section in all projects after 12 months 0 40 80 120 160 200 0 5 10 15 20 25 30 In te rf ac e Sh ea r St re ng th , p si FWD Center Deflection, mils Figure 4.9-1. Relationship between ISS and mean center deflection: all projects. 0 10 20 30 40 50 60 70 80 Range 0 20 40 60 80 100 120 140 160 180 F re qu en cy Interface Shear Strength, psi N = 320 µ = 76 psi = 32 psi Figure 4.10-1. Histogram of ISS measurements for all tested samples.

41 of service according to the procedures described in the Distress Identification Manual for the Long-Term Pavement Perfor mance Program (50). Surveyed distress types included rutting and surface cracking. The test sections in all projects performed satisfactorily with no presence of rutting and surface cracking, except a few sections in the Missouri project that exhibited low to moderate surface cracking with no rutting. Figure 4.10-2 presents the relationship between ISS test results and the number of surface cracks observed in the 33 test sections of all field projects. In the Missouri project, the number of cracks increased noticeably when the average ISS dropped below the minimum ISS threshold (1). 4.11 ISS Predictive Model Development An ISS predictive model was developed with the use of the 0-month service ISS test results. This model was intended for use by practitioners during the planning of overlay construc- tion activities to select tack coat type and application rate, given the project conditions. The field experimental program evaluated the effects of tack coat materials, application rates, and pavement surface textures on ISS for four types of pave- ment surfaces: milled HMA, new HMA, existing HMA, and PCC. For this purpose, field projects were selected in dif- ferent states, and specimens were cored from the evaluated test sections. The LISST device, a direct shear test setup, was used to characterize interface bonding of extracted core specimens (1). It was determined from the LISST results that interface bonding between pavement layers depends on several factors, including pavement surface type, tack coat material type, surface texture, and application rate. In an effort to identify the correlation with measured ISS results, physical properties of pavement surface type (i.e., MTD), resid- ual application rate, and characteristics of tack coat materials (i.e., penetration, softening point, useful temperature inter- val, rotational viscosity, and high temperature performance grade) were measured. (The useful temperature interval is defined here as the difference between the high temperature grading and the low temperature grading.) All these vari- ables were considered in the initial development of the pre- dictive model. 4.12 Data Description for Model Development As previously noted, the experimental program in NCHRP Project 09-40A consisted of 10 field projects in six states and included 33 in-service test sections. The 33 test conditions evaluated in the current project were not sufficient for model development and validation. To overcome this limitation, LISST data from the recently completed project, NCHRP Project 09-40 (1), were used to complement the data collected in NCHRP 09-40A. A description of both NCHRP Project 09-40 and NCHRP Project 09-40A is provided below. 4.12.1 NCHRP Project 09-40 Experimental Program As part of NCHRP Project 09-40 (1), full-scale test sections—including different types of tack coat materials and application rates between a new HMA overlay and several pavement surface types—were constructed at the Louisiana Transportation Research Center Pavement Research Facility. 0 40 80 120 160 0 2 4 6 8 10 12 In te rf ac e Sh ea r St re ng th , p si Number of Cracks LA MO FL TN NV OK Minimum Recommended ISS Threshold (1) Figure 4.10-2. Relationship between ISS and number of surface cracks: all projects.

42 The objective of the experimental program was to measure and to compare the ISS for different surface types, tack coat material types, and residual application rates. Table 4.12-1 presents the experimental factorial conducted in NCHRP Project 09-40. The variables and their ranges were carefully selected on the basis of a worldwide survey. Four types of pavement surfaces and four types of tack coat materials were evaluated; however, only the SS-1H tack coat material was used on new HMA, and two emulsion grades (SS-1H and SS-1) were used on milled HMA surface. Samples were collected from the test sections and were evaluated in the laboratory with the use of the LISST device. Details of this experimental program were presented elsewhere (1). 4.12.2 NCHRP Project 09-40A Experimental Program Table 4.12-2 presents the experimental factorial conducted in NCHRP 09-40A. Six types of tack coat materials—SS (SS-1H, CSS-1H, and SS-1) and nontracking RS (CBC-1H, CRS-1HBC, and NTSS-1HM)—were evaluated on four pave- ment surface types. However, CRS-1HBC and SS-1 tack coat materials were used only on existing HMA and milled HMA surfaces. ISS test results were analyzed in relation to the mea- sured residual application rates. Residual application rates measured in the field experimental program ranged from 0.01 to 0.08 gsy; see Table 4.12-2. 4.13 Model Development Methodology To develop the proposed predictive model, the measured ISS at 0 month of service (i.e., immediately after overlay construction) was selected as the dependent variable, and pavement type (i.e., HMA or PCC), surface texture depth, residual application rate, and tack coat rheological proper- ties were identified as potential independent variables. Sur- face type and surface texture were selected as two distinct variables because surface texture alone was not sufficient to distinguish between HMA and PCC surfaces. Table 4.13-1 presents the selected dependent and independent variables that were evaluated during model development. LISST data from NCHRP Project 09-40 and NCHRP Project 09-40A were utilized for both model development and validation purposes. The data comprised 50 test sections and were divided into 70% (i.e., 35 test sections) and 30% (i.e., 15 test sections) for model development and validation, respectively. To deter- mine the correlation between dependent and independent variables, the SAS was used to perform the statistical analyses (i.e., Pearson’s correlation, coefficient of determination, and general linear model procedure). Then, the stepwise variable selection method was used to identify potential independent variables relevant in predicting the dependent variable. After the linear regression model was developed, multicollinearity of independent variables was evaluated with the use of col- Variable Content Levels Pavement surface type Existing HMA, New HMA, Milled HMA, PCC 4 Tack coat material SS-1H, SS-1, CRS-1, NTSS-1HM 4 Residual application rate, gsy 0.03, 0.06 2 Number of replicates 3 3 Note: Several variables were partially evaluated according to the test factorial (1). Table 4.12-1. NCHRP Project 09-40 experimental factorial: model development. Variable Content Levels Pavement surface type Existing HMA, New HMA, Milled HMA, PCC 4 Tack coat material SS-1H, SS-1, CSS-1H, CRS-1HBC, CBC-1H, NTSS-1HM 6 Residual application rate, gsy 0.01-0.08 N/A Number of replicates 3 3 Table 4.12-2. NCHRP Project 09-40A experimental factorial: model development.

43 linearity diagnostics (i.e., condition index, variance inflation factor, and tolerance value). Finally, nonlinear regression analy- sis was performed with the Excel Solver tool to obtain the most accurate predictive model. Test data for all independent variables obtained from the evaluated test sections were eval- uated initially for potential correlation analysis; see Table C-9 in Appendix C. The following sections present two statistical approaches of model development, namely, multiple linear regression analysis and nonlinear regression analysis. 4.14 Multiple Linear Regression Analysis The objective of the multiple linear regression analysis was to model the relationship between the dependent variable (Y) and the independent variables (Xs). The multiple linear regression model was represented by the following equation: = β + β + + β. . . (6)0 1 1Y X XK K where β0 = regression coefficient for the intercept and βi = regression coefficients for the independent variables X1 through XK. The first step in developing the multiple linear regression model was selecting the appropriate independent variables. This task involved computing the pairwise correlation coeffi- cient between any two variables expected to be used in the model. Adding more variables in a multiple regression analy- sis may increase the coefficient of determination, r2, of the model, but it does not necessarily mean that the model will provide better prediction. This problem in regression analysis is called overfitting. Further, more independent variables will also increase the probability of including correlated inde- pendent variables. Not only can the independent variables be potentially related to the dependent variable, but they can be related to each other. This problem is called multicollinearity. In the development of the multiple linear regression model, overfitting and multicollinearity were evaluated. To avoid over- fitting and multicollinearity problems, a stepwise regression analysis was performed with the use of the SAS, including all independent variables to identify potential independent variables relevant in predicting the dependent variable. Stepwise regression analysis is a method of fitting all pos- sible regression models, where a variable is considered to be included or rejected from a set of independent variables based on predefined criterion (73). This method consists of varia- tions of two basic ideas, that is, forward selection and back- ward elimination. The forward selection method starts with no variable in the model and adds one variable at a time until either all variables are included or a stopping criterion is sat- isfied. The backward stepwise selection method starts with all independent variables in the model. The F-statistic is calculated for each independent variable. The least significant variables are removed from the model if they do not meet the significance level. This iterative procedure was repeated until no nonsignificant variables were left to be considered. In the stepwise regression analysis, a significance level of 0.25 was selected for a variable to be included in the model, and a variable had to be significant at the 0.15 level to remain in the model. This significance level was determined on the basis of Variable Type Variables Description Dependent ISS, psi Maximum interface shear stress Po te nt ia l I nd ep en de nt V ar ia bl es C at eg or ic al Pavement surface type (PT) HMA and PCC C on ti nu ou s Residual application rate (RES), gsy Measured in the field during construction MTD, mm Useful temperature interval (UTI), ºC Rheological properties of tack coat materials PG high temperature (HPG), ºC Penetration (PEN) at 25oC, 100g, 5 s, dmm Softening point (SFPT), ºC Rotational viscosity (RV) at 135°C, Pa.s Note: PT was selected as a discrete variable in the analysis on the basis of two clusters of pavement types that assume two values, 0 and 1, indicating whether the pavement surface type was PCC or HMA, respectively. Differences in the HMA pavement surface type are addressed by the MTD. Table 4.13-1. Selected variables in characterizing interface bonding.

44 the recommended levels in SAS for stepwise regression analy- sis procedure. Additionally, a linear regression analysis was performed considering all variables to compare the results with the stepwise selection method. 4.14.1 Pairwise Correlation Analysis To avoid overfitting and multicollinearity problems in the statistical predictive model, a pairwise correlation analy- sis was performed. Collinearity or multicollinearity can be problematic in regression analysis because linearly cor- related independent variables may increase the variance of the estimated regression coefficients; this factor makes the coefficients unstable and difficult to interpret (64). Iden- tification of this problem involves computing pairwise cor- relation coefficients between any two of all independent variables expected to be used in the model. The Pearson’s product–moment correlation coefficient or the linear cor- relation coefficient, r, is a measure of strength and direction of linear association between the selected variables. This correlation coefficient is a convenient index of strength of the linear relationship between two variables (65). The maxi- mum value of the correlation coefficient varies from -1 to +1; a value of zero indicates no relationship between two vari- ables. The sign of a correlation coefficient indicates positive or negative relationship between two variables. In general, a large absolute value of the correlation coefficient (i.e., greater than 0.9) indicates potential multicollinearity. The Pearson’s correlation coefficient, r, can be defined by the following equation: ∑ ∑ ∑ ∑ ∑ ∑ ∑( ) ( ) ( ) ( )( ) ( ) ( ) = − −  −  (7) 2 2 2 2 r n xy x y n x x n y y where x = x variable, y = y variable, and n = number of variables. Table 4.14-1 summarizes the Pearson correlation coef- ficients for all variables initially considered. As expected, rheological properties (i.e., PEN, SFPT, RV, HPG, and UTI) of tack coat materials were highly correlated with each other. Therefore, it was concluded that of the five rheological vari- ables, one or two rheological variables that were most sig- nificantly related to the others would be sufficient in the prediction model. 4.14.2 Correlation Between Dependent and Independent Variables The coefficient of determination, r2, the square of the Pearson correlation coefficient, r, was calculated for all inde- pendent variables. The r2-value is important because it pro- vides a measure of variation in the dependent variable that can be predicted from the independent variables. An ANOVA between the dependent variable, ISS, and the independent variables was conducted with the use of SAS. The ANOVA provided a p-value that helped to understand whether that independent variable was significant to the model. Indepen- dent variables that were not statistically significant in the model were removed. If, from the ANOVA analysis, a set of correlated variables were significant, then according to the pairwise correlation analysis, only one should be retained, and the selection can be made on the basis of the p-value. Table 4.14-2 summarizes the results of the ANOVA analysis. It was observed that four independent variables (i.e., PEN, SFPT, RV, and UTI) were not statistically significant and were, therefore, removed from the developed model. 4.14.3 Model Development A multiple linear regression analysis was conducted with the PROC REG procedure on the basis of the list of significant independent variables identified in the previous section. The following equation presents the multiple regres- sion model that was developed with the SAS software to predict ISS: VAR ISS RES MTD PEN SFPT RV HPG UTI ISS 1.00 0.76 0.66 -0.34 0.23 0.22 0.32 0.30 RES 0.76 1.00 0.34 -0.29 0.09 0.15 0.12 0.06 MTD 0.66 0.34 1.00 0.11 -0.13 -0.12 -0.08 0.05 PEN -0.34 -0.29 -0.11 1.00 -0.87 -0.82 -0.88 -0.77 SFPT 0.23 0.09 -0.13 -0.87 1.00 0.93 0.95 0.77 RV 0.22 0.15 -0.12 -0.82 0.93 1.00 0.90 0.73 HPG 0.32 0.12 -0.08 -0.88 0.95 0.90 1.00 0.88 UTI 0.30 0.06 0.05 -0.77 0.77 0.73 0.88 1.00 Table 4.14-1. Summary of Pearson correlation analysis.

45 = − + + + + ISS 69.5 17.9 * PT 897.4 *RES 27.7 * MTD 0.49*HPG (8) where ISS = predicted ISS, psi; PT = pavement surface type (i.e., 0 = PCC and 1 = HMA); MTD = in mm; RES = residual application rate, gsy; and HPG = high PG, °C. The range of values for parameters used for model devel- opment is as follows: • Residual application rate (RES) is between 0.02 to 0.06 gsy, • Mean texture depth (MTD) is between 0.06 to 2.2 mm, and • HPG is between 50 to 100 °C. Figure 4.14-1 illustrates the correlation between the measured and predicted ISS based on the developed model. This figure shows a good correlation with an r2 value of .85. A 95% confidence interval was constructed from the multiple linear regression model. Similarly, a 95% prediction interval, which indicated the range of values that was likely to contain the response value of a single new observation in the model, was developed. The root mean square error (RMSE) or root mean squared deviation was also calculated as a measure of the difference between the predicted and measured observa- tions, according to the following equation: ∑ ( )= −=RMSE 1 ˆ (9) 2 1n y yi ii n where RMSE = root mean square error, yi = measured value, ŷi = predicted value, and n = number of observations. The calculated RMSE was 8.9 psi. The range of measured ISS values used in model development was between 19 and 99 psi, a wide range encompassing both poorly bonded and strongly bonded tack coated interfaces. 4.14.4 Multicollinearity Test In multiple regression analysis, it is possible that there is no pairwise correlation, but a combination of independent Variable F-Value P-Value PT 62.24 <.0001 RES 98.17 <.0001 MTD 150.07 <.0001 PEN 1.43 0.2424 SFPT 1.31 0.2625 RV 1.18 0.2869 HPG 5.95 0.0218 UTI 0.31 0.5842 Table 4.14-2. Summary of ANOVA. Figure 4.14-1. Predicted ISS versus measured ISS (model development).

46 variables is correlated with some other combinations of independent variables. The variance inflation factor (VIF) is the most common statistic factor used to verify this prob- lem. VIF is a measure of inflation in the standard error (SE) associated with a particular weight due to multicollinearity. For example, a VIF of 8 implies that the SEs are larger by a factor of 8 than would otherwise be the case, if there were no intercorrelations between the predictor of inter- est and the remaining predictor variables included in the multiple regression analysis. Various recommendations for acceptable levels of VIF have been published in the litera- ture. A value of 5 or 10 has typically been recommended as the maximum level of VIF (66). Table 4.14-3 presents the multicollinearity test results. The VIF values for all independent variables were less than 5, which indicates no multicollinearity problem exists between and among the independent variables. 4.14.5 Model Validation As previously noted, 30% of the observations were used in the model validation phase as independent observations. Figure 4.14-2 presents the validation of the proposed linear model by comparing the measured and the predicted ISS val- ues. There was an acceptable correlation between the measured and the predicted ISS values with an r2 value of .80 and a RMSE value of 6.8 psi. The range of measured ISS values used for model validation was between 27 and 75 psi. As previously mentioned, the multiple linear regression model showed an r2 value of .85 with an RMSE value of 8.9 psi in the development phase. On the basis of the range of measured ISS (19 to 99 psi) and RMSE, the percentage error would range between 9% and 47%. The COV of the RMSE is 16.6% for the RMSE value of 8.9 psi, which is calculated by normalizing RMSE by the mean value of ISS (i.e., 53.6 psi) Variable SE t-Value P-Value Tolerance VIF Intercept 11.9 -5.85 <.0001 N/A N/A PT 3.9 4.60 <.0001 0.94 1.06 RES 142.8 6.28 <.0001 0.80 1.25 MTD 4.1 6.72 <.0001 0.77 1.29 HPG 0.13 3.78 .0007 0.98 1.03 Table 4.14-3. Multicollinearity test results. Figure 4.14-2. Predicted ISS versus measured ISS (model validation).

47 used in model development. This COV is deemed acceptable. However, if the minimum ISS value (i.e., 19 psi) were used to normalize RMSE, the obtained percentage error would be 47%, which was considered high and may not be acceptable. This situation motivated an evaluation of nonlinear regres- sion analysis to achieve better prediction accuracy. The fol- lowing section presents the development of the nonlinear regression model. 4.15 Nonlinear Regression Analysis Nonlinear regression analysis is a statistical technique that is used to describe a nonlinear relationship between dependent and independent variables. Nonlinear regression models are generally assumed parametric, where the model is described as a nonlinear equation. Nonlinear regression uses several algorithms including the Gauss–Newton, Marquardt– Levenberg, Nelder–Mead, and steepest descent methods (67). The principle of this method is the same as for linear regression analysis: to minimize the sum of squared error between data and the fitted values (68). An iterative non- linear least square fitting method was employed using the Excel Solver tool to develop the nonlinear regression model. The Solver tool uses an iteration protocol that is based on the robust and reliable generalized reduced gradient method. This method requires an initial estimate of the parameter values on the basis of prior experience or reasonable guess of the function used to fit the data. The iteration is repeated by changing the parameter values and functions until the minimum possible value of the sum of squared error is achieved. 4.15.1 Model Development After approximately 30 model forms were attempted, the following equation represents the most accurate nonlinear model form that was developed to predict the ISS: [ ]( )= × + + × ×ISS 3 2 MTD PT 8 HPG RES (10)3 where ISS = predicted ISS, psi; PT = pavement surface type (i.e., 0 = PCC and 1 = HMA); MTD = mean texture depth, mm; RES = residual application rate, gsy; and HPG = high performance grade, °C. The range of values for parameters used for model devel- opment is as follows: • Residual application rate is between 0.02 to 0.06 gsy, • MTD is between 0.06 to 2.2 mm, and • HPG is between 50 to 100 °C. Figure 4.15-1 illustrates the correlation between the pre- dicted and measured ISS based on the developed nonlinear regression model. The model showed an excellent r2 value Figure 4.15-1. Predicted ISS versus measured ISS (model development).

48 of .95 with an RMSE value of 4.8 psi. Recall that the range of measured ISS values used for model development was between 19 and 99 psi, representing both poorly bonded and strongly bonded tack-coated interface. 4.15.2 Model Validation Figure 4.15-2 presents validation of the proposed non- linear model by comparing the measured and predicted ISS values. The validation was performed with a different dataset that was independent from the one used for model develop- ment. There was a good correlation between the measured and predicted ISS values with an r2 value of .84 and a RMSE value of 6.2 psi. The range of measured ISS values used for model validation was between 27 and 75 psi. The nonlinear regression model showed an r2 value of .95 with an RMSE value of 4.8 psi in the development phase and an r2 of .84 with an RMSE of 6.2 psi in the validation phase. On the basis of the range of the measured ISS values and RMSE, the percentage error ranged from 5% to 25%, which was a considerable reduction compared to the linear model and was deemed acceptable. It was thus indicated that the nonlinear regression was a more suitable approach than linear regression, as better prediction accuracy was accomplished. Hence, the developed nonlinear model was selected as the recommended predictive model to estimate interface bond- ing during planning of overlay construction activities. 4.16 Illustrative Applications of the ISS Predictive Model During the planning of overlay construction activities, a contractor, a state agency, or both may be interested in selecting a tack coat material and application rate that would provide adequate ISS at the site. Tables 4.16-1 through 4.16-4 present applications of the developed nonlinear predic- tive model in estimating ISS values. Four MTD values were selected as default MTD values for different pavement sur- face types on the basis of the measured mean MTD values shown in Table 4.2-1. Five types of tack coat materials were selected on the basis of HPG, namely, high (NTSS-1HM), intermediate (CBC-1H, SS-1H and CSS-1H), and low (SS-1); the corresponding HPG values are set as default in Tables 4.16-1 through 4.16-4. Further, various residual application rates (RES)—low (0.035 gsy), medium (0.045 gsy), and high (0.055 gsy)—for each type of tack coat material and surface type are presented in Tables 4.16-1 through 4.16-4. The HPG tack coat material (NTSS-1HM) showed the highest inter- face bonding (i.e., 87 psi) for milled HMA surface type at the highest application rate (i.e., 0.055 gsy), while the low PG tack coat material (SS-1) showed the minimum interface bonding strength (i.e., 26 psi) for PCC pavement surface at the lowest application rate (i.e., 0.035 gsy). These estimated ISS values showed good agreement with the measured ISS values obtained from the current study. Figure 4.15-2. Predicted ISS versus measured ISS (model validation).

49 Tack Coat Type HPG RES Predicted ISS NTSS-1HM 84 0.055 87 84 0.045 77 84 0.035 67 CBC-1H/SS-1H/CSS-1H 72 0.055 79 72 0.045 71 72 0.035 62 SS-1 50 0.055 65 50 0.045 59 50 0.035 53 Note: MTD = 1.77 mm; PT = 1. Table 4.16-1. Illustrative applications of the ISS predictive model: milled HMA. Tack Coat Type HPG RES Predicted ISS NTSS-1HM 84 0.055 66 84 0.045 56 84 0.035 46 CBC-1H/SS-1H/CSS-1H 72 0.055 58 72 0.045 49 72 0.035 41 SS-1 50 0.055 43 50 0.045 37 50 0.035 31 Note: MTD = 0.91 mm, PT = 1. Table 4.16-2. Illustrative applications of the ISS predictive model: new HMA. Tack Coat Type HPG RES Predicted ISS NTSS-1HM 84 0.055 67 84 0.045 57 84 0.035 47 CBC-1H/SS-1H/CSS-1H 72 0.055 59 72 0.045 50 72 0.035 42 SS-1 50 0.055 44 50 0.045 38 50 0.035 32 Note: MTD = 0.97 mm, PT = 1. Table 4.16-3. Illustrative applications of the ISS predictive model: existing HMA. Tack Coat Type HPG RES Predicted ISS NTSS-1HM 84 0.055 60 84 0.045 50 84 0.035 40 CBC-1H/SS-1H/CSS-1H 72 0.055 52 72 0.045 44 72 0.035 35 SS-1 50 0.055 38 50 0.045 32 50 0.035 26 Note: MTD = 1.49 mm, PT = 0. Table 4.16-4. Illustrative applications of the ISS predictive model: PCC.